Class LPScheduler::FluentTracking

Defined in File lpscheduler.h

Nested Relationships

This class is a nested type of Class LPScheduler.

Class Documentation

class FluentTracking

A class to track the changes on a task numeric variable as the LP is built for the plan so far.

Public Types

enum fluent_status

An enum defining three states a numeric variable can be in when building the LP:

FS_NORMAL denotes that the variable should be treated normally, with effects and preconditions on the variable being included in the LP.
FS_IGNORE denotes that the variable is a metric tracking variable, to be ignored (i.e. effects on it should be omitted from the LP)
FS_ORDER_INDEPENDENT denotes that the variable is an order-independent metric-tracking variable, and effects on it are to update the variables FluentTracking::orderIndependentValueTerms and FluentTracking::orderIndependentValueConstant, for eventual inclusion into the objective function.

Values:

enumerator FS_NORMAL

enumerator FS_IGNORE

enumerator FS_ORDER_INDEPENDENT

Public Functions

inline FluentTracking(const double &initial)

Initial constructor: the variable takes the given value (that from the initial state), its status is set to FS_NORMAL, and no continuous numeric change is active.

Parameters:: initial – The value of the fluent in the initial state.

inline FluentTracking(): Default constructor (unfortunately needed to be able to create a vector of FluentTracking objects).

Public Members

fluent_status statusOfThisFluent: The status of this fluent. For metric-tracking variables, this is set to either FS_IGNORE or FS_ORDER_INDEPENDENT in the LPScheduler constructor, depending on whether the plan is to be scheduled to the metric. For all other variables, this is set to FS_NORMAL.

double postLastEffectValue: The value of the variable following the last effect. This is only defined if FluentTracking::lastEffectValueVariable = -1.

int lastEffectValueVariable: The LP variable (column) containing the value of the variable following the last effect to act upon it. If this takes the value -1, then the value is a constant, not an LP variable, stored in FluentTracking::postLastEffectValue.

int lastEffectTimestampVariable: The timestamp variable of the last action with an effect upon this variable. If this takes the value -1, then no such action exists (i.e. there have been no effects on the variable so far).

double activeGradient: The gradient of the active continuous numeric change acting upon the variable.

int activeGradientCount: How many actions are having a continuous numeric effect upon the variable. This is used to prevent rounding errors: if this value takes the value 0, then FluentTracking::activeGradient is assigned the value 0.

map<int, double> orderIndependentValueTerms: For metric-traking variables with order-independent effects, this map contains column—weight pairs that contribute to its value.

double orderIndependentValueConstant: For metric-tracking variables with order-independent effects, the sum of this constant value and the weighted sum of the LP columns given in FluentTracking::orderIndependentValueTerms give the value of the variable at the end of the plan.